- atrf-xmit.c (xfer_one): use atrf_interrupt_wait instead of
atrf_interrupt to pull the interrupt line
- atrf-xmit.c (xfer_one): removed delay-based wait to interrupt
- atrf-xmit.c (xfer_one): use atrf_interrupt_wait instead of reading
REG_IRQ_STATUS directly
- include/misctxrx.h (wait_for_interrupt), lib/misctxrx.c
(wait_for_interrupt): replaced sleep_us*timeout limiting mechanism
with a single timeout value in milliseconds
- lib/misctxrx.c (wait_for_interrupt): use atrf_interrupt_wait instead
of polling
- lib/misctxrx.c (wait_for_interrupt): cleaned up control flow
- lib/misctxrx.c (run, die, wait_for_interrupt): renamed variable "run"
to more specific "sigint"
- atrf-txrx/atrf-txrx.c (ping_rx, ping): pass timeout in milliseconds,
not deciseconds
- atrf-rssi/atrf-rssi.c (sweep), atrf-rssi/gui.c (sweep),
atrf-txrx/atrf-txrx.c (init_txrx, receive_message, receive_pcap,
receive, transmit, transmit_pattern, ping_tx, ping_rx),
atrf-xmit/atrf-xmit.c (init_tx, init_rx, xfer_one),
lib/cwtest.c (enter_test_mode_230, start_test_mode_231):
updated use of wait_for_interrupt
- atben.c (atben_interrupt_wait): loop until either the timeout has
been reached or an interrupt has occurred
- atben.c (atben_driver): added atben_interrupt_wait
- atusb-common.c (atusb_open): claim interface, so that we can do bulk
I/O without complaints from the kernel
- atusb-common.h (atusb_interrupt_wait), atusb-common.c: blocking read
for interrupt status byte on EP1
- atusb.c (atusb_driver), atusb-spi.c (atusb_spi_driver): provide the
interrupt_wait operation
- tools/lib/driver.h (struct atrf_driver): new driver operation
interrupt_wait
- include/atrf.h (atrf_interrupt_wait), atrf.c: new function to
wait for a transceiver interrupt without polling
- dirtpan.c: changed control byte from innovative 7 bit layout to the
more common 8 bits. My, haven't we had our morning coffee yet ?
- dirtpan.c: added explanation that the two highest bits of the
control byte have to be zero, to avoid conflicting with RFC4944
- include/cwtest.h (cw_test_needs_reset), lib/cwtest.c
(cw_test_needs_reset): new function to indicate all cases where the
transceiver needs to be reset (instead of using SLP_TR)
- lib/cwtest.c (cw_test_end): use cw_test_needs_reset instead of
open-coding the decision
- lib/cwtest.c (enter_test_mode_230, start_test_mode_231): always wait
for the PLL to lock. Contrary to the assumption in the previous
commit, we should always see this interrupt.
- atrf-path/atrf-path.c (sample): initialize the transceiver if we had
to reset it
- atrf-path/atrf-path.c (sample), atrf-path/atrf-path.c (do_half_sweep):
moved the tTR19 delay to "sweep", so that all branches share it and it
is taken only once
We seem to have a path where the PLL is already locked when we get
there. We thus never get an interrupt and hang.
Pending further investigation, just use a timeout that is longer
than the worst-case PLL lock time.
- atrf-xmit.c (xfer_one): also check that also the sender has no
pending interrupts before we start sending
- atrf-xmit.c (xfer_one): interrupt poll loop (now disabled) checked
the sender, not the receiver. oops.
- atrf-xmit.c (xfer_one): upon reception, check that the sender has
finished (IRQ_TRX_END) as well
- gui.c (old_term, restore_term, raw, get_key): raw and non-blocking
console input
- gui,c (gui): accept console input in addition to keypressed in the
SDL window
- atrf-gpio.c (usage): added newline after synopsis
- atrf-gpio.c (usage): description of option -p was missing
- atrf-gpio.c (main): invocation without patterns/commands is now an
error
- atrf-gpio.c: added section titles
- atrf-gpio.c (DEFAULT_DELAY_MS, reg_op, usage): the "delay" command
now accept an optional delay argument
- atrf-gpio.c (old_term, restore_term, raw, pass_fail): make standard
input non-blocking and raw, and poll for pass/fail input
- atrf-gpio.c (usage, main): new option -c to cycle through the
pattern sequence, waiting for pass/fail input
- atrf-gpio.c (reg_op, usage): new command "delay" to add a 10 ms
delay
- atrf-gpio.c (reg_op, usage): new command "frame" to write one
byte to the frame buffer
- atrf-gpio.c (reg_op, usage): new command "reset" to reset the
transceiver (via atrf_reset_rf)
- atrf-gpio.c (reg_op, usage): new command "slp_tr" to pulse SLP_TR
(via atrf_slp_tr)
- atrf-gpio.c (reg_op, usage): register read can now be followed by a
mask value
ATUSB can't use sleep mode because that would cut the AVR's clock.
The old SiLabs-based boards would not have that issue, but there,
the SLP_TR method is unproven. Besides, reset is as fast on USB.
This, while being slightly unsystematic, reduces the risk of confusing
"1" (one) and "l" (lower-case ell).
- tools/atrf-gpio/atben.c (do_atben), tools/atrf-gpio/atusb.c
(do_atusb): treat 'o' as an alias of 'l'
- tools/atrf-gpio/atrf-gpio.c (usage): list 'o' as alternative for 'l'
- atusb.c (name): clarified that PC3 is unconnected
- atusb.c (gpio): "dir" and "data" were swapped
- atusb.c (gpio): complain if ATUSB_GPIO returns the wrong amount of data
- atusb.c (dump, main): show the pin configuration as well
- atusb.c (dump): show "-" for values where we don't expect a specific
result
- atusb.c (dump): removed extra newline on mismatch
- Makefile: removed limitation to Ben and added host build
- atusb.c: atusb-specific acquisition and calculations
- atrf-xtal.h (do_atusb): interface to atusb.c
- atrf-xtal.c (atben, atusb, main): added switch to board-specific driver
- atrf-xtal.c (usage, main): added support for atusb
This means that we only depend on f32xbase for atusb-sil (defunct) and
cntr (needs updating), but not for any of the project's main parts.
- makefiles/Makefile.basic: copied over from
f32xbase/lib/Makefile.common
- makefiles/Makefile.basic (LDLIBS): removed inclusion of libusb
- tools/Makefile.common: use makefiles/Makefile.basic
- install/INSTALL-Ben, prod/doc/setup.hmac: removed f32xbase
dependency
- gui.c (segment, draw, gui): make caller provide color, to ease sharing
- gui.c (dump, gui): moved call to print_sweep to separate function
- gui.c (DUMP_RGBA, dumps, n_dumps, dump, draw_dumps, gui): record all
dumps and display them in the background
- sweep.h (print_sweep), atrf-path.c (print_sweep): make global, for
sharing with gui.c
- gui.c (gui): dump the sweep results to standard output if "D" is
pressed
- atrf-xtal.c (eval): return the value instead of printing it
- atrf-xtal.c (usage, main): new option -b for reporting the deviation
from a base count
- atrf-xtal.c (eval): new option -p to report an error for deviations
outside the specified bound
Note that the PHY is still half-duplex, but we don't need to insist
on the entire packet reception/transmission to finish before letting
the peer have its say.
- dirtpan.c (open_tun): if a command is given, set the environment
variable ITF to the interface name and system(3) the command
- dirtpan.c (usage, main): accept an interface configuration command as
the 4th command-line argument
Note that we still waste at least one perfectly good frame. However,
instead of implementing some fancy arbitration, it's better if we
just make the whole thing bidirectional in the future.
The IEEE 802.15.4 kernel stack doesn't implement ACKs yet and prints
a warning if we request them. Since dirtpan has its own ACKs, we
just disable IEEE 802.15.4 ACKs for now.
- dirtpan.c (debug_ip): only dump local packets if debug > 1
- dirtpan.c (debug_dirt): dump packet content only if debug > 1
- dirtpan.c (debug_dirt): made packet content dump more readable
- dirtpan.c (debug_dirt): added terse debug output for debug == 1
- dirtpan.c (usage): doubling -d now has a meaning
- usbwait.c (need_removal, wait_for_usb): if need_removal is set, wait
for the device to be removed before trying to find it
- usbwait.c (usage, main): new option -r to set need_removal
- usbwait/usbwait.c: update the USB device tree until the specified
device is detected
- usbwait/Makefile: build usbwait (host only)
- Makefile (DIRS, BEN_DIRS, upload): separated subdirectories with tools
that can run on the Ben from those that can't
- Makefile (DIRS): added usbwait/
- usbopen.c (open_usb): determine vendor and product objective outside
the loop
- usbopen.c (open_usb): check vendor and product only if non-zero
- usbopen.c (bad_id, parse_usb_id): vendor and product can now be
omitted
- tools/include/usbopen.h (usb_rescan), tools/lib/usbopen.c (initialized,
initialize, usb_rescan): usb_rescan forces a re-initialization of
libusb and a new scan of the USB device tree
- include/atrf.h (atrf_driver_spec), lib/atrf.c (atrf_driver_spec):
new function to retrieve the local or remote/final driver spec
- lib/atrf.c (struct atrf_dsc, atrf_open, atrf_close): record the local
driver spec
- lib/driver.h (struct atrf_driver): new driver function "driver_spec"
to retrieve the driver spec
- lib/atnet.c (struct atnet_dsc, atnet_open, atnet_close): maintain a
cache for the driver spec
- lib/atnet.c (atnet_driver_spec, atnet_driver): added support for the
"driver_spec" function
- atrf-proxy/PROTOCOL, atrf-proxy/atrf-proxy.c (cmd_zero): added command
SPEC to retrieve the (final) driver spec
- atrf-id/atrf-id.c (usage, main): added option -s to retrieve the
driver spec. One -s retrieves the local spec, -s -s the remote/final.
- Makefile (DIRS): added atrf-xmit
- atrf-xmit/Makefile, atrf-xmit/atrf-xmit.c: new utility atrf-xmit to
rapidly send packets from one device to another and analyze the outcome
- gui.c (disc, indicate): moved drawing of disc to separate function
- gui.c (triangle, up, down): draw an up/down triangle
- gui.c (indicate): use up/down triangle for over/under
- atrf-path.c (do_sweeps): return a pass/fail/undecided value
- gui.h (gui), gui.c (gui): return a pass/fail/undecided value
- gui.c (gui): return pass/fail when P or F is pressed; exit
unconditionally if Q is pressed
- atrf-path.c (main): according to the decision of "gui" or "do_sweeps",
print "#PASS", "#FAIL", or nothing
- gui.c (avg2y, segment): moved mapping from measurement to y position
to separate function, to allow for sharing
- gui.c (LIMIT_RGBA, draw_limit, gui): draw the limits
- sweep.h (MIN_DIFF, MAX_DIFF, struct sweep): added min/max profile
- sweep.h (do_sweep), atrf-path.c (do_half_sweep, do_sweep): compare
result against limits and return pass/fail decision
- sweep.h (N_CHAN), atrf-path.c (do_half_sweep, do_sweeps), gui.c
(N_CHAN): declare number of channels in one central place instead of
scattering it all around the program
- atrf-path.c (do_read_profile, read_profile, usage, main): new option -P
to read a min/max profile
- gui.c (indicate, main): moved indicator to separate function and
improved blink logic
- gui.c (OVER_RGBA, UNDER_RGBA, indicate, main): change color to indicate
pass/fail
atrf-path -T +0.5 -g ... 10 10 time:
before: 0.32-0.33 s
after: 0.27-0.29 s (88%)
- atrf-path.c (sample): force cw test setup if never done before
- atrf-path.c (do_half_sweep): only request cw test setup if sweeping
both offsets
atrf-path -T +0.5 -g ... 10 10 time:
before (both offsets) 0.65 s
after (one offset) 0.32-0.33 s (50%)
- atrf-path.c (usage, main): new option -T to specify which offset to
sweep (default: sweep both)
- atrf-path.c (do_half_sweep): only sample points with the desired offset
- atrf-path.c (print_sweep): only prints points we've sweeped
- gui.c (segment, draw): inverted flag logic from "have_last" to "first"
- gui.c (draw): only plot points we've sweeped
- gui.c (gui): pass the offset selection to "draw"
atrf-path -g ... 10 10 time:
before 1.92 s
after 0.65 s (34%)
- atrf-path.c (do_sweep): separate -0.5 MHz and +0.5 MHz sweep
- atrf-path.c (sample, do_sweep): set up cw test mode only twice per
sweep (once for each offset), and simply resume in all other cases
atrf-path -g ... 10 10 time is now stable at 1.92 s
- atrf-path.c (sample): commented-out init sequence needed for AT86RF230
- atrf-path.c (sample, do_sweep): set TX channel only once per +/-0.5 MHz
pair
- atrf-path.c (sample, do_sweep): "sample" no longer needs the channel
argument
- atrf-path.c (sample, main): do TX init only once
- lib/cwtest.c (enter_test_mode_231, prepare_test_mode_231,
start_test_mode_231, cw_test_begin): broke enter_test_mode_231 into
a slow setup component and a fast start/resume component
- include/cwtest.h (cw_test_resume), lib/cwtest.c (cw_test_resume):
resume transmission in a previously set up test mode
atrf-path -g ... 10 10 time:
before 2.08 s
after 1.92-1.93 s (92-93%)
- atrf-path.c (set_channel, init_common): moved setting of the channel
to separate function, for better granularity
- atrf-path.c (init_common, init_tx, init_rx, sample, do_sweep): the
init_* functions no longer set the channel; use set_channel
- atrf-path.c (main): initialize the receiver only once
- atrf-path.c (usage, main): new option -g to invoke the GUI
- atrf-path.c (usage, main): in GUI mode, the number of sweeps is optional
(default: infinite)
- atrf-path.c (main), gui.h (gui), gui.c (gui): exit after the specified
number of sweeps
The transceiver reset takes quite a while. It seems that also going into
sleep mode has the effect of stopping cw test mode and it's much faster.
This has only been tried on the AT86RF231. For AT86RF230, we therefore
still use the reset.
Performance improvement:
Measurement: atrf-path sweep time in GUI mode, averaging over 10 samples.
Running on PC with remote ATBEN transceivers.
before 7.36-7.68 s
after 2.08 s (27-28% of previous time)
- gui.h, gui.c: plot sweep results with SDL_gfx
- Makefile: added target-specific variables for SDL and SDL_gfx
- atrf-path.c (main): invoke the GUI if the number of sweeps is zero
